[SOLVED] Ryzen 9 5900x suddenly overheating with no apparent thermal throttling.

[mecheverriasu]

For the past two weeks I've had, all of a sudden, higher than usual CPU temps. My 5900x is now going all the way up to 90°C while gaming and under high stress scenarios (Benchmarking, Unreal Engine), sometimes even just using Firefox. I first noticed this issue while playing Fortnite, because the stuttering became very noticeable and after a while I saw the CPU was down to 1.8 GHz. Since that day, I've remounted the 240mm AIO and changed the thermal paste three times, and my CPU would still go up to 90°C in certain scenarios. However, I'm still getting around 8000-8500 points in Cinebench R20, which is a fairly normal value as I've seen from various sources online (no overclocking, -28 curve offset on all cores), and HWInfo has shown no Thermal Throttling, even when the CPU went severely down on it's clock. I've checked the AIO, and the pump seems to be working (one hot tube and one much colder tube, with slight vibration).

Today, while playing League of Legends, there was apparent throttling again. So I closed every program I had open, and after waiting for maybe 20 minutes, the CPU was still at 90°C, and any attempt to use a web browser or any other program would throttle it down to 0.5 GHz. I turned off my system and, for some reason, RGB lights stayed on. Afraid of having the CPU be still pulling power with no active cooling or safety software/firmware to shut it off, I unplugged the power for a few minutes. After a while, I turned on the system again and went to the BIOS. CPU still went up to 70-90°C and while looking for solutions on my phone, it finally cooled down after like two minutes. I turned the system on and everything seems to be functioning normally as of right now, but I'm entirely clueless as to what is actually happening with my system.

This is a picture of my system throttling with HWInfo

https://imgur.com/lCSF3sK

And this is me still getting normal benchmark scores after turning my pc on again (didn't have hwinfo open, only afterburner to show temps).

https://imgur.com/5hMfatp

My system:
CPU: AMD Ryzen 9 5900x
MB: Aorus Elite B450
AIO: Lian Li Galahad 240
Thermal Paste: Thermal Grizzly Kryonaut
RAM: Corsair Vengeance 3200MHz 2x32GB

 

[Phaaze88]

I've checked the AIO, and the pump seems to be working (one hot tube and one much colder tube...)

Even if the pump is 'working', that's not the symptom of good flow.
"I've checked the AIO, and the pump seems to be working (the tubes feel about the same...)"
^That is.

RMA the cooler. There's nothing you can do for it at this point.
If you are past the warranty window, then you've got to buy another cooler.

 

[mecheverriasu]

UPDATE:

Two nights ago I decided to reset the CMOS, considering that maybe something was wrong with the BIOS settings. To do this, I used the jumper in the motherboard (I unplugged the power before that). When I tried to power on the PC again, it wouldn't post, and the motherboard LED's showed DRAM issue.

Yesterday, I tested my ram sticks in my brother's PC and they would post and launch windows, but only twice, after that, it would not post again and his motherboard would show CPU led.

So at this point I decided to try an old 8gb 2133MHz stick in my system, and surprised, it launched windows with no issues. I decided to test my two original sticks in all slots. One of them wouldn't post at all, and the other one would always post but only launch Windows 1 of 10 times (it'd get stuck right when the loading dots were on screen, after the BIOS opening period was over). Both together didn't work at all.

I decided to format my drive and reinstall windows. The not-working-at-all stick was still unusable, but the other one (the 1 out of 10 times) now decided to work without apparent issues, but only with XMP disabled, at 2133MHz instead of 3200MHz. But the most interesting part, is that my CPU is not overheating anymore (which was the main reason of this thread in the first place).

Right now I'm using my only usable stick without XMP, because with XMP enabled the system doesn't post at all which requires me to clear the CMOS again (I've done this twice to confirm that XMP is really not working). Is there something that could've gone wrong the first time I cleared the CMOS two nights ago that somehow damaged my RAM sticks?

 

[Phaaze88]

considering that maybe something was wrong with the BIOS settings.

Did you, or anyone else, make any manual changes to bios settings before all this? [Ignoring XMP]


Since that kit also fails in your brother's PC, I'd suggest contacting Corsair about a warranty replacement.

 

[mecheverriasu]

Did you, or anyone else, make any manual changes to bios settings before all this? [Ignoring XMP]


Since that kit also fails in your brother's PC, I'd suggest contacting Corsair about a warranty replacement.

PBO was set to advanced by Ryzen Master when I used the curve optimizer, but this was after the CPU overheat issues started. Also some other stuff like disabling USB power when off and enabling resizable bar, but not anything else I remember, that could involve RAM or CPU. I'm just gonna use the warranty on the RAM as you mentioned, hopefully the store will accept it, as they are less than a year old.

Still, the fact that both sticks with XMP were working fine before clearing the CMOS makes me believe I did something wrong. At least the CPU is now working fine, apparently.

 

[Phaaze88]

If the store doesn't accept it, you should go through Corsair.

The only thing that came to mind was bad manual bios settings, like SOC voltage, which is sometimes suggested for high(er) memory OCs. Users will sometimes take it too far, and wreck their cpu's Internal Memory Controller; losing the ability to run at memory speeds it once could. @drea.drechsler
But since the memory crapped itself in your brother's PC, I disregarded the possibility.

 

[DSzymborski]

Yeah, it sounds like multiple things are going on.

As noted above, having the tubes being very different temperatures is an extremely bad sign. The temperature delta between the tube coming from the CPU and the tube going back down should only be a couple degrees.

 

[mecheverriasu]

Thanks for the answers so far.

I've decided to go through store warranty for the RAM. However, this process will take some time (up to 15 days) according to the store (I'm from Peru so RMA through Corsair directly would be difficult and impractical). I still need my pc for work, so I decided to buy another almost identical kit from Corsair as well, which is working fine so far. If I get a replacement from the store I'd probably sell it anyway to recover some of the cost.

I also decided to buy an air cooler (Noctua NH D15) through Amazon, which is arriving next week. Honestly I'm kind of done with AIO's. The performance may be good but it just seems they're so much more prone to defects and I just don't want to deal with that. I'll give you an update after I install it and test it. Regarding my current AIO, I believe the warranty has already expired, so I might as well just open it myself and see if I can fix it with online guides.

My cpu is no longer overheating on regular use conditions. As I mentioned before, it used to reach 80°C+ during regular use and gaming, and so far only reaches 90°C during full load, and up to 70°C while gaming. This should not happen with a 240mm AIO anyway, but we'll see after I install the air cooler. I've been suspecting, as mentioned by Phaaze88, the IMC could be degrading and my motherboard under-reporting power consumption, but I don't know how to test that. My new RAM is working fine at 3200MHz and max 1.013V on the SOC (according to HWInfo that is, not real measure), so maybe that´s not the case.

 

[mecheverriasu]

Update:

I just got the Noctua NH-D15 and the cpu is literally 20°C cooler under full load. So, for now, I guess the AIO was actually screwed. Thanks for the help, hopefully I won't have anymore issues in the close future.